Brett Adcock has transformed Figure AI from a startup burning $1 million per month into a manufacturing powerhouse capable of producing one humanoid robot every hour. In just four months, the company scaled production from one unit per day to 55 Figure 03 robots weekly, with an 80% first-pass yield rate that demonstrates serious manufacturing discipline. This 24x production increase positions Figure competitively in an expanding humanoid robot market while Adcock simultaneously manages two other high-intensity ventures: Cover, a school safety company using NASA-derived imaging technology, and Hark, an AI lab focused on personalized intelligence hardware.

How Is Figure AI Achieving Industrial-Scale Robot Production?

Figure's BotQ facility in California now operates like a precision manufacturing plant rather than a prototype workshop. Over 350 Figure 03 units have shipped, each surviving 80 functional tests including robotic squats and jogging sessions. The company has achieved battery production yields of 99.3% while custom actuators roll off dedicated production lines. This infrastructure includes over 150 networked workstations running custom manufacturing software, representing a fundamental shift from garage-built hardware to industrial-grade production.

The manufacturing success stems from Adcock's philosophy of total vertical integration. Rather than relying on external vendors, Figure designs and manufactures everything in-house, from motors and rotors to battery packs and sensors. As Adcock explained, this approach prevents the company from being "left at the mercy of some vendor". The long-term goal is even more ambitious: removing nearly all Chinese components from the Figure 03 by summer 2026 while scaling toward 1 million units per year.

As Adcock

What Makes Figure's Latest Robot Upgrade Genuinely Impressive?

The Helix System 0 upgrade represents a breakthrough in robot navigation that addresses one of robotics' most persistent challenges: getting machines to move through unpredictable environments without human intervention. The system combines stereo vision with proprioception, essentially giving robots spatial awareness similar to how humans understand their position in space. Trained through reinforcement learning in simulation, the model transfers directly to real-world robots without requiring prior mapping of specific locations.

This sim-to-real transfer is the kind of technical achievement that robotics engineers have pursued for years. Previously, robots would struggle with stairs and uneven terrain, often failing in ways that seemed almost comical. Now, the Helix AI integration enables zero-shot stair navigation, meaning robots can handle stairs they've never encountered before without tumbling down like expensive metal pinballs.

How Is Adcock Managing Three Separate High-Intensity Ventures?

Adcock's approach to scaling multiple companies simultaneously reflects an extreme commitment to what he calls playing the game "11 out of 10." He has cut out almost all social activities, including golf trips and dinners with old friends, to focus exclusively on his family and his companies. This intensity extends across three distinct business areas:

• Figure AI: The flagship humanoid robotics company targeting 12,000 units annually and scaling toward 1 million units per year, with a current valuation of $39 billion.
• Cover: A school safety company that acquired intellectual property from NASA's Jet Propulsion Lab involving terahertz imaging radar technology originally used to detect explosives in war zones, now being repurposed to detect concealed weapons in K-12 schools with beta deployments planned for end of 2026.
• Hark: An AI lab and hardware company launched in late 2025 focused on "personalized intelligence" and designing devices that move beyond the "20-year-old interface" of phones and laptops, recently securing a full datacenter of NVIDIA B200 GPUs for model training.

During Figure's first four months of operation, Adcock self-funded the venture through a $1 million monthly burn rate to hire a 40-person team of what he calls "best athletes". This early sprint established the foundation for the manufacturing race now underway at BotQ.

Why Did Adcock End Figure's Partnership with OpenAI?

One of the most revealing aspects of Adcock's strategy involves his decision to dissolve Figure's partnership with OpenAI, a collaboration that initially seemed like a landmark alliance between large language model (LLM) expertise and physical robotics. Adcock stated bluntly that his internal team, largely comprised of veterans from Google DeepMind, was "running circles" around OpenAI's robotics efforts.

"We were just way better at this, so I fired them," Adcock stated, adding that the relationship became a strategic liability once it became clear Figure was effectively teaching OpenAI how to approach robot learning while receiving diminishing returns.

Brett Adcock, CEO at Figure AI

This pivot allowed Figure to double down on its own "omni-model" architecture, which powers the Helix 02 system to compute torque directly from pixels without third-party interference. The decision reflects Adcock's broader philosophy of maintaining complete control over the technology stack rather than depending on external AI providers.

What's the Timeline for Figure's Production Targets?

Figure's production roadmap reveals both ambition and a grounded understanding of manufacturing realities. The company hit record production levels in March 2026 and plans to triple that output by May. The current target of 12,000 units annually represents a significant milestone, but the four-year goal of 100,000 units per year demonstrates the scale Adcock envisions.

These machines will initially serve industrial and commercial applications rather than consumer markets. The proof of concept lives in consistent delivery rather than press releases, as the real test comes when field deployments generate enough data to refine the Helix AI model through edge-case failures and real-world scenarios. If Figure maintains this production cadence while improving functionality, the company will have solved what many consider the hardest part of the humanoid robot equation: building them reliably at scale.

What Challenges Remain for Figure's Expansion?

Despite the manufacturing breakthroughs, Adcock acknowledges that the path to "general robotics," a machine that can do anything a human can, remains a "fun house of problems" that has yet to be fully solved. The company is also navigating the fallout from a whistleblower lawsuit alleging that the pursuit of speed has at times compromised safety.

The real competitive advantage will emerge from the combination of manufacturing scale and continuous AI improvement. With over 350 robots already deployed and production ramping to one unit per hour, Figure is generating the kind of real-world data that can refine the Helix AI system in ways that simulation alone cannot achieve. The company's bet is that sheer volume of data combined with vertical control over the entire stack will be enough to crack the code of physical artificial intelligence.